Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
2018
Abstract
Aim: To develop a combined predictive model for preterm and term pre-eclampsia (PE) during the first tri-
mester of pregnancy.
Methods: This investigation was a nested case–control study in singleton pregnancies at the Maternal-Fetal
Medicine Unit, University of Chile Hospital. A priori risks for preterm and term PE were calculated by multi-
variate logistic regression analyses. Biophysical markers were log10-transformed and expressed as multiples
of the median. A multivariate logistic regression analysis was used to estimate a combined predictive model
of preterm and term PE. Detection rates at different cut-off points were determined by a receiver operator
curve analysis of a posteriori risks.
Results: First trimester mean arterial pressure and uterine artery Doppler pulsatility index were significantly
higher in women who develop PE than in the unaffected group. The detection rate of preterm PE based on
maternal characteristics and biophysical markers was 72% at a 10% false-positive rate, corresponding to a
cut-off risk of 1 in 50. The detection rate for term PE was 30% at a 10% false-positive rate.
Conclusion: Preterm PE can be predicted by a combination of maternal characteristics and biophysical
markers. However, first trimester screening is less valuable for term PE.
Key words: Doppler, first trimester, mean arterial pressure, prediction, pre-eclampsia, uterine artery.
positive rate of 10%, the detection rates (DR) vary, the International Society for the Study of Hyperten-
depending on the population, from 47% to 96%.8,14 sion in Pregnancy.16 Preterm PE (p-PE) and term PE
Interestingly, the incorporation of the mean arterial (t-PE) were defined as individuals with PE who deliv-
pressure (MAP) at the first trimester screening has ered before and after 37 weeks of gestation,
been demonstrated to improve the DR of these respectively.
models. Small for gestational age (SGA) was defined as a
The aim of this study was to develop a first trimes- birthweight <10th percentile, according to local stan-
ter combined predictive model for preterm and term dards.17 Fetal growth restriction (FGR) was defined as
PE that includes maternal history and biophysical SGA with altered Doppler, SGA delivered before
markers. 37 weeks of gestation due to fetal or maternal condi-
tion, or a birthweight less than 3rd percentile, irre-
spective of Doppler status or gestational age at
Methods birth.18
was stored, and the pulsatility index (PI) was calcu- obtained regression formula was used to determine
lated.20 The mean PI of the left and right UtAD was the individual risk, in the same manner as a priori risk
calculated and stored. estimation. Finally, the DR at different a posteriori risk
cut-offs was estimated.
Study outcomes For group comparison, a P value less than 0.05 was
The primary outcome was the diagnosis of p-PE with considered significant.
or without FGR. The secondary outcome was the
diagnosis of t-PE.
Results
Statistical analysis
All statistical analyses were performed with STATA During the study period, we included 1812 patients.
v14.2 software (Statacorp). For continuous variables, Of these, 56 were excluded for spontaneous miscar-
Student’s t-test or the Mann–Whitney U test was used riage (30), stillbirth (8), neonatal death related to
and expressed as the mean (standard deviation) or major congenital defect (4) and live birth with a major
median (interquartile range [IQR]) for parametric and congenital defect (14). After exclusion, 1756 singleton
nonparametric distributions, respectively. Categorical pregnancies were included in the final analysis. Forty-
variables were compared with Pearson’s chi-square nine (2.8%) patients developed PE, and 29 of them
test and expressed as n (%). For the analysis of predic- were delivered before 37 weeks of gestation. Maternal
tive variables, a Bayes theorem approach was used. characteristics and the obstetric history of the study
population are presented in Table 1. Patients who
A priori risk subsequently developed PE were older and heavier
First, maternal characteristics were explored with a and had higher rates of smoking and chrHT than
univariate logistic regression analysis to determine patients with unaffected pregnancies. There were no
the odds ratios (OR) of variables significantly associ- differences in ethnicity, with a predominance of the
ated with the primary outcome. Second, a multivari- white Hispanic population in both groups. Twenty-
ate logistic regression analysis was applied, using a nine percent of patients in the PE group had a history
backward stepwise elimination method. Variables of PE in a previous pregnancy, compared to 2.5% of
with a P value less than 0.05 were retained in the final patients in the unaffected group (P < 0.0001).
model. Finally, to determine the individual risk, odds At screening, pregnancies that developed PE pre-
were calculated as follows: Odds = ey. Then, risk was sented a significantly higher SBP and DBP, with a
estimated as follows: Odds/(1 + Odds). The a priori MAP of 85 mmHg (78.8–93.3) and 80 mmHg
risk was log10-transformed before including in the a (73.3–86.3) in the PE and unaffected groups, respec-
posteriori risk model. tively (Table 2). The mean UtAD was also higher in
the PE group, with an almost threefold higher rate of
Biophysical predictor markers pregnancies with a PI above the 90th percentile. Peri-
First, MAP and UtAD were log10-transformed to natal outcomes are presented in Table 2.
make their distribution Gaussian. Second, a multivari-
ate linear regression was performed to identify which A priori risk
variables were significant predictors of log10MAP and A univariate logistic regression model was performed
log10UtAD in the unaffected group. The obtained to identify the maternal characteristic risk factors for
regression was used to determine multiples of the p-PE and t-PE. Regarding p-PE risk factors, chrHT
median (MoM) of both markers. For missing values of (OR = 13), systemic lupus erythematosus (SLE)
UtAD measurements in 14 controls, a multiple impu- (OR = 8), previous history of PE (OR = 27), SGA
tation analysis of missing values was performed. (OR = 12) and spontaneous preterm delivery
(OR = 8) presented higher risk. However, after a mul-
A posteriori risk tivariate analysis, only chrHT, SLE and a previous
To determine the a posteriori risk, a multivariate logis- history of PE were retained in the model (Table 3).
tic regression was applied, combining the a priori For t-PE, only BMI and smoking habit were associ-
risk(log10), MoM-MAP and MoM-UtAD. A receiver ated with a modest but significantly higher risk.
operator curve analysis was performed to determine The a priori risk for p-PE was obtained with the fol-
the DR at a fixed false-positive rate of 10%. The lowing formula:
y ¼ − 5:354954 + maternal age × 0:0218775 comparison to unaffected pregnancies (unaffected
+ ð1:534937 ½if chrHTÞ + ð2:026845 ½if SLE pregnancies 1.001 [0.983–1.017]; t-PE 1.005
[0.983–1.020]; and p-PE 1.015 [0.998–1.027]; P = 0.02).
+ ð0 ½if nulliparousÞ
In addition, in the unaffected pregnancies, the mean
– ð0:1196494 ½if parous without previous PEÞ UtAD PIlog10 was better predicted with the following
+ ð2:754261 ½if parous with previous PEÞ formula:
The a priori risk for t-PE was obtained with the fol- y ¼ 0:4190377 – ð0:0025075 × CRLÞ
lowing formula:
– 0:0029433 × maternal age years
+ ð0 ½if nulliparous + 0:0015571
y ¼ − 6:942792 + ð0:0832792 × BMIÞ
½if parous without previous PE
+ 1:254454 if smoking habit
+ 0:0794263 ½if parous with previous PE
Combined predictive model for PE curve [AUC]: 0.817 [0.725–0.908]). Following the inclu-
For p-PE, at a fixed false-positive rate of 10%, the DR sion of MoM-MAP and MoM-UtAD, the DR of the
based on a priori risklog10 was 62.1% (area under the model went up to 72.4% (AUC: 0.890 [0.837–0.955];
Table 3 Odds ratios for significant maternal characteristics after univariate and multivariate logistic regression analysis
Maternal characteristics Univariate regression Multivariate regression
OR (95% CI) P value OR (95% CI) P value
Preterm PE
Chronic hypertension 13.3 (5.78–30.44) <0.0001 4.9 (1.78–13.41) 0.002
SLE 8.4 (1.82–38.34) 0.006 7.5 (1.13–49.13) 0.037
Previous PE
Nulliparous Reference — Reference —
Parous no previous PE 0.99 (0.38–2.58) 0.98 0.98 (0.37–2.59) 0.97
Parous with previous PE 27.2 (10.55–69.97) <0.0001 17.2 (5.96–49.53) <0.0001
Previous SGA
Nulliparous Reference — Excluded† —
Parous no previous SGA 2.0 (0.88–4.66) 0.10
Parous with previous SGA 12.2 (2.39–61.91) 0.003
Previous sPTD
Nulliparous Reference — Excluded† —
Parous no previous sPTD 1.8 (0.76–4.19) 0.18
Parous with previous sPTD 8.0 (2.52–25.04) <0.0001
Term pre-eclampsia
Maternal weight (kg) 1.03 (1.00–1.06) 0.029 Excluded† —
Body mass index (kg/m2) 1.08 (1.01–1.17) 0.036 1.1 (1.01–1.17) 0.035
Smoking habit 3.4 (1.23–9.58) 0.018 3.5 (1.25–9.81) 0.017
†Variable excluded after multivariate regression analysis. CI, confidence interval; OR, odds ratio; PE, pre-eclampsia; SGA, small for gesta-
tional age; SLE, systemic lupus erythematosus; sPTD, spontaneous preterm delivery.
Figure 1 Boxplot of multiples of the median (MoM) for mean arterial pressure (MAP) and uterine artery Doppler (UtAD) at first
trimester scan. (a) MoM of MAP; (b) MoM of UtAD pulsatility index. PE, preeclampsia. *P value < 0.05 versus unaffected.
Figure 3). The following formula represented the best MAP and MoM-UtAD were not significantly associ-
predictive model for p-PE: ated with t-PE and were therefore not included in the
final model.
y ¼ − 5:117278 + 2:777415 × a priori risk log10 Table 4 shows the DR and false-positive rates of dif-
ferent a posteriori risk cut-offs for p-PE. At a cut-off risk
+ ð3:487746 × MoM − MAPÞ of 1 in 100, 27.2% of pregnancies were classified as
+ ð2:158282 × MoM − UtADÞ high-risk but with a 25.7% false-positive rate. To reduce
the false-positive rate without significantly affecting the
DR, a cut-off risk of 1 in 50 might be considered more
With regard to t-PE, at a fixed false-positive rate of appropriate for our population, due to the fact that
10%, the DR based on a priori risklog10 was 30.0% only 7.8% were classified as high-risk pregnancies with
(AUC: 0.690 [0.580–0.799]). However, both MoM- a 72% DR and only a 10.4% false-positive rate.
However, Wright D et al. described that in parous outcomes and contacting the patients. We also thank
patients, a priori risk detected 61.7% of p-PE at a 10% the research midwife, Mrs Cristina Aleuanlli, for her
false-positive rate, which is similar to our findings.11 continuous help and energy in the development of
Regarding the performance of combined models, in clinical research at our department. This study was
a recent study, O’Gorman et al. screened 35 948 sin- funded by Fondo Nacional de Desarrollo Científico y
gleton pregnancies at 11 + 0 to 13 + 6 weeks. They Tecnológico (FONDECYT) Registry No. 1130668.
demonstrated that by combining maternal characteris- FONDECYT was not involved in the study design,
tics, MoM-UtAD and MoM-MAP, the DR of p-PE was patient recruitment or data analysis.
70% (95% confidence interval 64–75) at a 10% fixed
false-positive rate.21 The inclusion of PLGF into the
predictive model marginally improved the DR to
75%. Moreover, our group published that adding
Disclosure
PLGF into a predictive model can only increase the
None declared.
DR in our population by three percentile points.14
Based on these findings and the eventual increase
in the cost by including biochemical markers in the References
universal screening in our population, we believe that
a model that includes maternal characteristics and 1. Steegers EAP, von Dadelszen P, Duvekot JJ, Pijnenborg R.
Pre-eclampsia. Lancet 2010; 376: 631–644.
biophysical markers has enough strength to detect the 2. Donoso E. Tras el cumplimiento del 5 Objetivo del Milenio:
vast majority of women who will benefit from taking Mortalidad materna en Chile, 2013. Rev Chil Obstet Ginecol
aspirin before 16 weeks of gestation. 2015; 80: 431–433.
Finally, from a clinical perspective, the algorithm 3. Bujold E, Roberge S, Lacasse Y et al. Prevention of pre-
eclampsia and intrauterine growth restriction with aspirin
developed by the Fetal Medicine Foundation in a ran-
started in early pregnancy: A meta-analysis. Obstet Gynecol
domized controlled trial demonstrated a 62% reduc- 2010; 116: 402–414.
tion of p-PE when 150 mg of aspirin was started at 4. Roberge S, Villa P, Nicolaides K et al. Early administration
11–14 weeks.5 However, a subanalysis demonstrated of low-dose aspirin for the prevention of preterm and term
that this benefit is not observed in chronically hyper- preeclampsia: A systematic review and meta-analysis. Fetal
tensive patients.23 In the abovementioned randomized Diagn Ther 2012; 31: 141–146.
5. Rolnik DL, Wright D, Poon LC et al. Aspirin versus placebo
study, a cut-off risk of 1 in 100 was considered the in pregnancies at high risk for preterm preeclampsia. N Engl
inclusion criteria, with values associated with a 10% J Med 2017; 377: 613–622.
screen-positive rate. However, if a similar cut-off 6. Askie LM, Duley L, Henderson-Smart DJ, Stewart LA. Anti-
point is used, more than 20% of our population platelet agents for prevention of pre-eclampsia: A meta-
should be classified using the screen-positive rate, analysis of individual patient data. Lancet 2007; 369:
1791–1798.
which might reflect an inadequate overuse of aspirin. 7. Parra M, Rodrigo R, Barja P et al. Screening test for pre-
Therefore, a cut-off risk of 1 in 50 should be more eclampsia through assessment of uteroplacental blood flow
suitable in our screening program to obtain a screen- and biochemical markers of oxidative stress and endothelial
positive rate of approximately 10%. dysfunction. Am J Obstet Gynecol 2005; 193: 1486–1491.
8. Poon LCY, Syngelaki A, Akolekar R, Lai J, Nicolaides KH.
In conclusion, we demonstrated that the use of a
Combined screening for preeclampsia and small for gesta-
first trimester model for the prediction of p-PE devel- tional age at 11–13 weeks. Fetal Diagn Ther 2013; 33: 16–27.
oped with a local population, could obtain similar DR 9. Poon LCY, Kametas NA, Maiz N, Akolekar R, Kypros H.
than international reports. However, t-PE might be First-trimester prediction of hypertensive disorders. Hyper-
unlikely to be predicted properly during the first tri- tension 2009; 53: 812–818.
mester of pregnancy. Local risk cut-offs should be 10. Akolekar R, Syngelaki A, Poon L, Wright D, Nicolaides KH.
Competing risks model in early screening for preeclampsia
selected before the implementation of preventive by biophysical and biochemical markers. Fetal Diagn Ther
strategies in a high-risk group. 2013; 33: 8–15.
11. Wright D, Syngelaki A, Akolekar R, Poon LC,
Nicolaides KH. Competing risks model in screening for pre-
eclampsia by maternal characteristics and medical history.
Acknowledgments Am J Obstet Gynecol 2015; 213: 62.e1–62.e10.
12. Poon LC, Nicolaides KH. First-trimester maternal factors
We thank the administrative staff of the Fetal Medi- and biomarker screening for preeclampsia. Prenat Diagn
cine Unit for their effort in obtaining perinatal 2014; 34: 618–627.
13. Scazzocchio E, Figueras F, Crispi F et al. Performance of a based management protocol. Fetal Diagn Ther 2014; 36:
first-trimester screening of preeclampsia in a routine care 86–98.
low-risk setting. Am J Obstet Gynecol 2013; 208: 203. 19. Verdecchia P, Angeli F, Mancia G et al. How can we use the
e1–203.e10. results of ambulatory blood pressure monitoring in clinical
14. Parra-Cordero M, Rodrigo R, Barja P et al. Prediction of early practice? Hypertension 2016; 11: 102–107.
and late pre-eclampsia from maternal characteristics, uterine 20. Jurkovic D, Jauniaux E, Kurjak A, Hustin J, Campbell S,
artery Doppler and markers of vasculogenesis during first Nicolaides KH. Transvaginal color Doppler assessment of
trimester of pregnancy. Ultrasound Obstet Gynecol 2013; 41: the uteroplacental circulation in early pregnancy. Obstet
538–544. Gynecol 1991; 77: 365–369.
15. Robinson HP, Fleming JE. A critical evaluation of sonar 21. O’Gorman N, Wright D, Syngelaki A et al. Competing risks
‘crown-rump length’ measurements. Br J Obstet Gynaecol model in screening for preeclampsia by maternal factors and
1975; 82: 702–710. biomarkers at 11–13 weeks gestation. Am J Obstet Gynecol
16. Tranquilli AL, Dekker G, Magee L et al. The classification, 2016; 214: 103.e1–103.e12.
diagnosis and management of the hypertensive disorders of 22. Eyheramendy S, Martinez FI, Manevy F, Vial C, Repetto GM.
pregnancy: A revised statement from the ISSHP. Pregnancy Genetic structure characterization of Chileans reflects histori-
Hypertens 2014; 4: 97–104. cal immigration patterns. Nat Commun 2015; 6: 1–10.
17. Milad M, Fabres J, Aspilaga CA. Recomendación sobre Cur- 23. Poon LC, Wright D, Rolnik DL et al. Aspirin for evidence-
vas de Crecimiento Intrauterino. Rev Chil Pediatr 2010; 81: based preeclampsia prevention trial: Effect of aspirin in pre-
264–274. vention of preterm preeclampsia in subgroups of women
18. Figueras F, Gratacós E. Update on the diagnosis and classifi- according to their characteristics and medical and obstetrical
cation of fetal growth restriction and proposal of a stage- history. Am J Obstet Gynecol 2017; 217: 585.e1–585.e5.